Re: PSPP-BUG: Re: [bugs #12035] AGGREGATE gets its sums wrong!

[Ben Pfaff]

John Darrington <address@hidden> writes:

> On closer inspection, this bug is not a problem with AGGREGATE, but
> with the DATA LIST.
>
>
>
> So far as I can tell, the offending code is: 
>
> RCS file: /cvsroot/pspp/pspp/src/data-in.c,v
> retrieving revision 1.12
> diff -b -w -r1.12 data-in.c
> 218,219d217
> <   else if (!got_dot)
> <     exponent -= i->format.d;
>
> I can't see any good reason for these lines.  

By way of explanation, here is what the SPSS manual says:

Implied Decimal Positions

      · For fixed-format data, decimal positions can be coded in
         the data or implied by the format. If decimal positions
         are implied but are not entered in the data, the program
         inter- prets the rightmost digits in each value as the
         decimal digits. A coded decimal point in a value
         overrides the number of implied decimal places. For
         example, (DOLLAR,2) specifies two decimal positions. The
         value 123 is interpreted as 1.23; however, the value
         12.3 is interpreted as 12.3 because the coded decimal
         position overrides the number of implied decimal
         positions.

Here is what the PSPP manual says:

       In addition, implied decimal places can be specified in parentheses
    after the column numbers.  As an example, suppose that a data file has a
    field in which the characters `1234' should be interpreted as having
    the value 12.34.  Then this field has two implied decimal places, and
    the corresponding specification would be `(2)'.  If a field that has
    implied decimal places contains a decimal point, then the implied
    decimal places are not applied.

It looks like the PSPP manual needs to say something similar
about FORTRAN style.  I will fix up the documentation.

> When I remove them, the problem goes away, and it passes all
> tests except:

The problem is, I believe, that we're applying implied decimal
places to free-format data, but it is only applicable for
fixed-format data.

If you want to fix this, go ahead; otherwise, I will do it when I
am done with the expression changes, which I still plan to commit
by the end of the month.
-- 
On Perl: "It's as if H.P. Lovecraft, returned from the dead and speaking by
seance to Larry Wall, designed a language both elegant and terrifying for his
Elder Things to write programs in, and forgot that the Shoggoths didn't turn
out quite so well in the long run." --Matt Olson